Merge pull request #2105 from stevvooe/algorithm-own-file

digest: cleanup digester and verifier creation
This commit is contained in:
Derek McGowan 2016-12-15 16:55:37 -08:00 committed by GitHub
commit 729b8c5b91
9 changed files with 211 additions and 188 deletions

145
digest/algorithm.go Normal file
View File

@ -0,0 +1,145 @@
package digest
import (
"crypto"
"fmt"
"hash"
"io"
)
// Algorithm identifies and implementation of a digester by an identifier.
// Note the that this defines both the hash algorithm used and the string
// encoding.
type Algorithm string
// supported digest types
const (
SHA256 Algorithm = "sha256" // sha256 with hex encoding
SHA384 Algorithm = "sha384" // sha384 with hex encoding
SHA512 Algorithm = "sha512" // sha512 with hex encoding
// Canonical is the primary digest algorithm used with the distribution
// project. Other digests may be used but this one is the primary storage
// digest.
Canonical = SHA256
)
var (
// TODO(stevvooe): Follow the pattern of the standard crypto package for
// registration of digests. Effectively, we are a registerable set and
// common symbol access.
// algorithms maps values to hash.Hash implementations. Other algorithms
// may be available but they cannot be calculated by the digest package.
algorithms = map[Algorithm]crypto.Hash{
SHA256: crypto.SHA256,
SHA384: crypto.SHA384,
SHA512: crypto.SHA512,
}
)
// Available returns true if the digest type is available for use. If this
// returns false, New and Hash will return nil.
func (a Algorithm) Available() bool {
h, ok := algorithms[a]
if !ok {
return false
}
// check availability of the hash, as well
return h.Available()
}
func (a Algorithm) String() string {
return string(a)
}
// Size returns number of bytes returned by the hash.
func (a Algorithm) Size() int {
h, ok := algorithms[a]
if !ok {
return 0
}
return h.Size()
}
// Set implemented to allow use of Algorithm as a command line flag.
func (a *Algorithm) Set(value string) error {
if value == "" {
*a = Canonical
} else {
// just do a type conversion, support is queried with Available.
*a = Algorithm(value)
}
return nil
}
// New returns a new digester for the specified algorithm. If the algorithm
// does not have a digester implementation, nil will be returned. This can be
// checked by calling Available before calling New.
func (a Algorithm) Digester() Digester {
return &digester{
alg: a,
hash: a.Hash(),
}
}
// New is deprecated. Use Algorithm.Digester.
func (a Algorithm) New() Digester {
return a.Digester()
}
// Hash returns a new hash as used by the algorithm. If not available, the
// method will panic. Check Algorithm.Available() before calling.
func (a Algorithm) Hash() hash.Hash {
if !a.Available() {
// Empty algorithm string is invalid
if a == "" {
panic(fmt.Sprintf("empty digest algorithm, validate before calling Algorithm.Hash()"))
}
// NOTE(stevvooe): A missing hash is usually a programming error that
// must be resolved at compile time. We don't import in the digest
// package to allow users to choose their hash implementation (such as
// when using stevvooe/resumable or a hardware accelerated package).
//
// Applications that may want to resolve the hash at runtime should
// call Algorithm.Available before call Algorithm.Hash().
panic(fmt.Sprintf("%v not available (make sure it is imported)", a))
}
return algorithms[a].New()
}
// FromReader returns the digest of the reader using the algorithm.
func (a Algorithm) FromReader(rd io.Reader) (Digest, error) {
digester := a.New()
if _, err := io.Copy(digester.Hash(), rd); err != nil {
return "", err
}
return digester.Digest(), nil
}
// FromBytes digests the input and returns a Digest.
func (a Algorithm) FromBytes(p []byte) Digest {
digester := a.New()
if _, err := digester.Hash().Write(p); err != nil {
// Writes to a Hash should never fail. None of the existing
// hash implementations in the stdlib or hashes vendored
// here can return errors from Write. Having a panic in this
// condition instead of having FromBytes return an error value
// avoids unnecessary error handling paths in all callers.
panic("write to hash function returned error: " + err.Error())
}
return digester.Digest()
}
// FromString digests the string input and returns a Digest.
func (a Algorithm) FromString(s string) Digest {
return a.FromBytes([]byte(s))
}

View File

@ -108,16 +108,17 @@ func (d Digest) Validate() error {
return ErrDigestInvalidFormat return ErrDigestInvalidFormat
} }
switch algorithm := Algorithm(s[:i]); algorithm { algorithm := Algorithm(s[:i])
case SHA256, SHA384, SHA512: if !algorithm.Available() {
if algorithm.Size()*2 != len(s[i+1:]) {
return ErrDigestInvalidLength
}
break
default:
return ErrDigestUnsupported return ErrDigestUnsupported
} }
// Digests much always be hex-encoded, ensuring that their hex portion will
// always be size*2
if algorithm.Size()*2 != len(s[i+1:]) {
return ErrDigestInvalidLength
}
return nil return nil
} }
@ -127,6 +128,15 @@ func (d Digest) Algorithm() Algorithm {
return Algorithm(d[:d.sepIndex()]) return Algorithm(d[:d.sepIndex()])
} }
// Verifier returns a writer object that can be used to verify a stream of
// content against the digest. If the digest is invalid, the method will panic.
func (d Digest) Verifier() Verifier {
return hashVerifier{
hash: d.Algorithm().Hash(),
digest: d,
}
}
// Hex returns the hex digest portion of the digest. This will panic if the // Hex returns the hex digest portion of the digest. This will panic if the
// underlying digest is not in a valid format. // underlying digest is not in a valid format.
func (d Digest) Hex() string { func (d Digest) Hex() string {
@ -141,7 +151,7 @@ func (d Digest) sepIndex() int {
i := strings.Index(string(d), ":") i := strings.Index(string(d), ":")
if i < 0 { if i < 0 {
panic("could not find ':' in digest: " + d) panic(fmt.Sprintf("no ':' separator in digest %q", d))
} }
return i return i

View File

@ -1,6 +1,8 @@
package digest package digest
import ( import (
_ "crypto/sha256"
_ "crypto/sha512"
"testing" "testing"
) )

View File

@ -1,141 +1,6 @@
package digest package digest
import ( import "hash"
"crypto"
"fmt"
"hash"
"io"
)
// Algorithm identifies and implementation of a digester by an identifier.
// Note the that this defines both the hash algorithm used and the string
// encoding.
type Algorithm string
// supported digest types
const (
SHA256 Algorithm = "sha256" // sha256 with hex encoding
SHA384 Algorithm = "sha384" // sha384 with hex encoding
SHA512 Algorithm = "sha512" // sha512 with hex encoding
// Canonical is the primary digest algorithm used with the distribution
// project. Other digests may be used but this one is the primary storage
// digest.
Canonical = SHA256
)
var (
// TODO(stevvooe): Follow the pattern of the standard crypto package for
// registration of digests. Effectively, we are a registerable set and
// common symbol access.
// algorithms maps values to hash.Hash implementations. Other algorithms
// may be available but they cannot be calculated by the digest package.
algorithms = map[Algorithm]crypto.Hash{
SHA256: crypto.SHA256,
SHA384: crypto.SHA384,
SHA512: crypto.SHA512,
}
)
// Available returns true if the digest type is available for use. If this
// returns false, New and Hash will return nil.
func (a Algorithm) Available() bool {
h, ok := algorithms[a]
if !ok {
return false
}
// check availability of the hash, as well
return h.Available()
}
func (a Algorithm) String() string {
return string(a)
}
// Size returns number of bytes returned by the hash.
func (a Algorithm) Size() int {
h, ok := algorithms[a]
if !ok {
return 0
}
return h.Size()
}
// Set implemented to allow use of Algorithm as a command line flag.
func (a *Algorithm) Set(value string) error {
if value == "" {
*a = Canonical
} else {
// just do a type conversion, support is queried with Available.
*a = Algorithm(value)
}
return nil
}
// New returns a new digester for the specified algorithm. If the algorithm
// does not have a digester implementation, nil will be returned. This can be
// checked by calling Available before calling New.
func (a Algorithm) New() Digester {
return &digester{
alg: a,
hash: a.Hash(),
}
}
// Hash returns a new hash as used by the algorithm. If not available, the
// method will panic. Check Algorithm.Available() before calling.
func (a Algorithm) Hash() hash.Hash {
if !a.Available() {
// NOTE(stevvooe): A missing hash is usually a programming error that
// must be resolved at compile time. We don't import in the digest
// package to allow users to choose their hash implementation (such as
// when using stevvooe/resumable or a hardware accelerated package).
//
// Applications that may want to resolve the hash at runtime should
// call Algorithm.Available before call Algorithm.Hash().
panic(fmt.Sprintf("%v not available (make sure it is imported)", a))
}
return algorithms[a].New()
}
// FromReader returns the digest of the reader using the algorithm.
func (a Algorithm) FromReader(rd io.Reader) (Digest, error) {
digester := a.New()
if _, err := io.Copy(digester.Hash(), rd); err != nil {
return "", err
}
return digester.Digest(), nil
}
// FromBytes digests the input and returns a Digest.
func (a Algorithm) FromBytes(p []byte) Digest {
digester := a.New()
if _, err := digester.Hash().Write(p); err != nil {
// Writes to a Hash should never fail. None of the existing
// hash implementations in the stdlib or hashes vendored
// here can return errors from Write. Having a panic in this
// condition instead of having FromBytes return an error value
// avoids unnecessary error handling paths in all callers.
panic("write to hash function returned error: " + err.Error())
}
return digester.Digest()
}
// FromString digests the string input and returns a Digest.
func (a Algorithm) FromString(s string) Digest {
return a.FromBytes([]byte(s))
}
// TODO(stevvooe): Allow resolution of verifiers using the digest type and
// this registration system.
// Digester calculates the digest of written data. Writes should go directly // Digester calculates the digest of written data. Writes should go directly
// to the return value of Hash, while calling Digest will return the current // to the return value of Hash, while calling Digest will return the current

View File

@ -17,17 +17,9 @@ type Verifier interface {
Verified() bool Verified() bool
} }
// NewDigestVerifier returns a verifier that compares the written bytes // NewDigestVerifier is deprecated. Please use Digest.Verifier.
// against a passed in digest.
func NewDigestVerifier(d Digest) (Verifier, error) { func NewDigestVerifier(d Digest) (Verifier, error) {
if err := d.Validate(); err != nil { return d.Verifier(), nil
return nil, err
}
return hashVerifier{
hash: d.Algorithm().Hash(),
digest: d,
}, nil
} }
type hashVerifier struct { type hashVerifier struct {

View File

@ -4,6 +4,7 @@ import (
"bytes" "bytes"
"crypto/rand" "crypto/rand"
"io" "io"
"reflect"
"testing" "testing"
) )
@ -12,10 +13,7 @@ func TestDigestVerifier(t *testing.T) {
rand.Read(p) rand.Read(p)
digest := FromBytes(p) digest := FromBytes(p)
verifier, err := NewDigestVerifier(digest) verifier := digest.Verifier()
if err != nil {
t.Fatalf("unexpected error getting digest verifier: %s", err)
}
io.Copy(verifier, bytes.NewReader(p)) io.Copy(verifier, bytes.NewReader(p))
@ -27,23 +25,42 @@ func TestDigestVerifier(t *testing.T) {
// TestVerifierUnsupportedDigest ensures that unsupported digest validation is // TestVerifierUnsupportedDigest ensures that unsupported digest validation is
// flowing through verifier creation. // flowing through verifier creation.
func TestVerifierUnsupportedDigest(t *testing.T) { func TestVerifierUnsupportedDigest(t *testing.T) {
unsupported := Digest("bean:0123456789abcdef") for _, testcase := range []struct {
Name string
Digest Digest
Expected interface{} // expected panic target
}{
{
Name: "Empty",
Digest: "",
Expected: "no ':' separator in digest \"\"",
},
{
Name: "EmptyAlg",
Digest: ":",
Expected: "empty digest algorithm, validate before calling Algorithm.Hash()",
},
{
Name: "Unsupported",
Digest: Digest("bean:0123456789abcdef"),
Expected: "bean not available (make sure it is imported)",
},
{
Name: "Garbage",
Digest: Digest("sha256-garbage:pure"),
Expected: "sha256-garbage not available (make sure it is imported)",
},
} {
t.Run(testcase.Name, func(t *testing.T) {
expected := testcase.Expected
defer func() {
recovered := recover()
if !reflect.DeepEqual(recovered, expected) {
t.Fatalf("unexpected recover: %v != %v", recovered, expected)
}
}()
_, err := NewDigestVerifier(unsupported) _ = testcase.Digest.Verifier()
if err == nil { })
t.Fatalf("expected error when creating verifier")
}
if err != ErrDigestUnsupported {
t.Fatalf("incorrect error for unsupported digest: %v", err)
} }
} }
// TODO(stevvooe): Add benchmarks to measure bytes/second throughput for
// DigestVerifier.
//
// The relevant benchmark for comparison can be run with the following
// commands:
//
// go test -bench . crypto/sha1
//

View File

@ -552,7 +552,7 @@ func testBlobAPI(t *testing.T, env *testEnv, args blobArgs) *testEnv {
}) })
// Verify the body // Verify the body
verifier, err := digest.NewDigestVerifier(layerDigest) verifier, err := layerDigest.Verifier()
if err != nil { if err != nil {
t.Fatalf("unexpected error getting digest verifier: %s", err) t.Fatalf("unexpected error getting digest verifier: %s", err)
} }

View File

@ -235,11 +235,7 @@ func (bw *blobWriter) validateBlob(ctx context.Context, desc distribution.Descri
// guarantee, so this may be defensive. // guarantee, so this may be defensive.
if !verified { if !verified {
digester := digest.Canonical.New() digester := digest.Canonical.New()
verifier := desc.Digest.Verifier()
digestVerifier, err := digest.NewDigestVerifier(desc.Digest)
if err != nil {
return distribution.Descriptor{}, err
}
// Read the file from the backend driver and validate it. // Read the file from the backend driver and validate it.
fr, err := newFileReader(ctx, bw.driver, bw.path, desc.Size) fr, err := newFileReader(ctx, bw.driver, bw.path, desc.Size)
@ -250,12 +246,12 @@ func (bw *blobWriter) validateBlob(ctx context.Context, desc distribution.Descri
tr := io.TeeReader(fr, digester.Hash()) tr := io.TeeReader(fr, digester.Hash())
if _, err := io.Copy(digestVerifier, tr); err != nil { if _, err := io.Copy(verifier, tr); err != nil {
return distribution.Descriptor{}, err return distribution.Descriptor{}, err
} }
canonical = digester.Digest() canonical = digester.Digest()
verified = digestVerifier.Verified() verified = verifier.Verified()
} }
} }

View File

@ -41,11 +41,7 @@ func TestSimpleRead(t *testing.T) {
t.Fatalf("error allocating file reader: %v", err) t.Fatalf("error allocating file reader: %v", err)
} }
verifier, err := digest.NewDigestVerifier(dgst) verifier := dgst.Verifier()
if err != nil {
t.Fatalf("error getting digest verifier: %s", err)
}
io.Copy(verifier, fr) io.Copy(verifier, fr)
if !verifier.Verified() { if !verifier.Verified() {